issn 1045-6333 harvard · cognitive biases affect the operation of legal rules, but also the legal...
TRANSCRIPT
ISSN 1045-6333
HARVARD JOHN M. OLIN CENTER FOR LAW, ECONOMICS, AND BUSINESS
THE SUCCESS AND SURVIVAL OF CAUTIOUS OPTIMISM: LEGAL RULES
AND ENDOGENOUS PERCEPTIONS IN PRE-TRIAL SETTLEMENT NEGOTIATIONS
Oren Bar-Gill
Discussion Paper No. 375
07/2002
Harvard Law School Cambridge, MA 02138
This paper can be downloaded without charge from:
The Harvard John M. Olin Discussion Paper Series: http://www.law.harvard.edu/programs/olin_center/
JEL Classification: B52, C15, C61, C78, D62, D89, K41
The Success and Survival of Cautious Optimism
Legal Rules and Endogenous Perceptions in Pre -Trial Settlement Negotiations
Oren Bar-Gill∗
July, 2002
Abstract
Litigants are unrealistically optimistic with regard to the probability of prevailing at trial. This systematic bias is well documented, and has been often invoked to explain breakdowns in pre-trial settlement negotiations. Contrary to existing models that allow for optimism as an exogenous assumption, the present study derives this cognitive bias endogenously. It thus provides a theoretical foundation for optimism in litigation. Quasi-evolutionary forces - market pressure (in the market for legal services) and imitation processes – are shown to favor cautiously optimistic litigants. Moreover, the endogenous optimism model enables an examination of the factors that determine the magnitude of the optimism bias. In particular, it is shown that the legal environment influences the equilibrium level of optimism. Focusing on rules for the allocation of litigation costs, the American rule induces a higher level of optimism, as compared to the British rule. This finding qualifies the conventional wisdom regarding the advantage of the American rule in fostering settlements. Finally, the present analysis is offered as an illustration of a broader theme, that the law can play an important role in determining the types and magnitudes of prevailing cognitive biases. The identification, characterization and analysis of this perception-shaping role of legal institutions are a novelty of the present study. Behavioral law and economics is revealed as a two-way, rather than a one-way street. Not only do cognitive biases affect the operation of lega l rules, but also the legal rules themselves influence the types and magnitudes of observed biases.
Keywords: Litigation, Pre-Trail Settlement Negotiations, Optimism.
∗ Harvard University, The Society of Fellows; and Harvard Law School, The Olin Center for Law,
Economics and Business. I wish to thank Jennifer Arlen, Lucian Bebchuk, Bernard Black, Eddie Dekel, Chaim Fershtman, Steven Shavell and seminar and conference participants at the 2002 American Law and Economics Association conference, Harvard Law School and Tel-Aviv University for their valuable comments and suggestions. Financial support from the Olin Center for Law, Economics and Business at Harvard Law School and from the Cegla Institute for Comparative and Private International Law at Tel-Aviv University is gratefully acknowledged.
1
The Success and Survival of Cautious Optimism
Legal Rules and Endogenous Perceptions in Pre-Trial Settlement Negotiations
Oren Bar-Gill∗
© 2002 Oren Bar-Gill. All rights reserved.
1. Introduction
Over 90% of all legal proceedings end up outside the courtroom, via settlement.
Still in a large number of cases settlement negotiations fail, and considerable amounts of
litigation costs are incurred at trial. These facts have lead to an extensive literature
regarding the prerequisites for settlement and the terms of settlements reached through
pre-trial negotiations (for recent surveys, see Daughety (2000) and Hay and Spier
(1998)).
The law and economics literature, having recognized that without information
problems parties will always settle1, has turned to models of uncertainty and asymmetric
information in attempt to explain the failure to reach settlement (see, for example, Cooter
et al. (1992), Priest and Klein (1984), Bebchuk (1984), Reinganum and Wilde (1986),
Nalebuff (1987), Spier (1992,1994) and Farmer and Pecorino (1996)). The main insight
suggests, that when uncertainty or asymmetric information induce a sufficiently large gap ∗ Harvard University, The Society of Fellows; and Harvard Law School, The Olin Center for Law,
Economics and Business. I wish to thank Jennifer Arlen, Lucian Bebchuk, Bernard Black, Eddie Dekel, Chaim Fershtman, Steven Shavell and seminar and conference participants at the 2002 American Law and Economics Association conference, Harvard Law School and Tel-Aviv University for their valuable comments and suggestions. Financial support from the Olin Center for Law, Economics and Business at Harvard Law School and from the Cegla Institute for Comparative and Private International Law at Tel-Aviv University is gratefully acknowledged.
1 A settlement saves litigation costs, thereby ensuring the existence of a settlement range.
2
between the (net) expected gain at trial, as perceived by the plaintiff, and the (net)
expected liability at trial, as perceived by the defendant, the parties will fail to reach a
settlement.2
However, uncertainty and asymmetric information are not the only explanation for
the common breakdown of settlement negotiations. Indeed, in many cases the relevant
information, at large, is known to both parties, and still settlement is not always reached.
Legal scholars and practitioners often attribute such breakdowns in settlement
negotiations to an optimism bias shared by many lawyers and litigants (see Birke and Fox
(1999) and the references cited therein3; see also Kaplow and Shavell (2000) (section
5.2.1) and Lowenstein et al. (1993)). The law and economics literature includes incidental
discussions of the implications of optimism in litigation and settlement contexts (see, e.g.,
Shavell (1982) and Lowenstein et al. (1993)). The origins of the optimism bias, however,
were never considered. Optimism, when recognized, was taken to be an exogenously
imposed impediment to settlement, not susceptible to analytical examination. 4
The present study sets out to derive optimism endogenously, rather than assume its
existence exogenously. It is shown that quasi-evolutionary forces operating in the
litigation environment induce “cautious optimism”. When lawyers and litigants with a
2 Of course, settlement will not be reached only if the plaintiff believes that his (net) expected gain at trial
is greater than the (net) expected liability that the defendant believes she faces at trial. If the defendant believes her (net) expected liability to be greater than (net) gain as perceived by the plaintiff, a settlement will surely be reached.
3 Birke and Fox (1999) provide a comprehensive survey covering the relevant psychological literature on the optimism bias, as well as specific studies, which prove the prevalence of a systematic bias towards optimism in legal settings. For an account of optimism in bargaining - see Bottom and Paese (1999). Bottom and Paese (1999) also summarize the experimental literature, which demonstrates the persistence of optimism in various settings.
4 A main problem with the exogenous view of optimism is that it precludes any analysis of the determinants of the optimism bias as well as any attempt to influence the level of the bias. Birke and Fox (1999) is a recent exception to this common view. These authors focus on practical measures, which can mitigate optimism as well as several other psychological biases that impede upon settlement negotiations.
3
variety of bias levels, ranging from pessimism to extreme optimism, interact with each
other in the pre-trial setting, only the bias levels that earn the highest average payoff will
survive in the long run. The evolution of lawyers' optimism has the greatest intuitive
appeal. A lawyer with a systematic bias that leads to below average profits will receive
little or no business and will eventually be forced out of the market (compare: Alchian
(1950), Friedman (1953), Camerer (1992) and Dutta and Radner (1999)).5 Alternatively,
successful lawyers will pass on their bias levels, through the training process of legal
interns, judicial clerks and junior associates6 and through their general influence on the
legal profession (see Bowles (1998), p. 82: "The cultural transmission process translates
economic well-being, exposure to roll models, and other influences into replication of
traits, and thus intervenes between payoffs and replication.").7
But, why will these quasi-evolutionary forces select “cautious optimism”? Why not
select pessimism, realism or more extreme levels of optimism? The answer builds on an
analysis of the two effects of optimism in the litigation and settlement context. First, as
previously recognized, optimism may prevent settlement. This force pushes towards less 5 On the cases that this lawyer does get, she will lose money. Thus, even if her below-average
performance is not made public and thus does not affect her caseload, she will eventually be driven out of the market.
6 Optimism may also be selected through the law firm promotion process. A cautiously optimistic associate will do better on average, and will thus have greater chances of being made partner.
7 The evolutionary methodology links the present study to papers, which attempt to explain seemingly irrational behavior as the result of an evolutionary selection process. See, for example, Kyle and Wang (1997), Waldman (1994) and Heifetz and Spiegel (1999) (Heifetz and Spiegel also demonstrate the evolutionary stability of cautious optimism albeit in an entirely different setting). More generally, our analysis relates to the growing literature on endogenous preferences. Prominent examples of this growing literature are: Frank (1988), Guth and Yaari (1992), Fershtman and Weiss (1997, 1998), Huck and Oechssler (1998), Bester and Guth (1998) and Dekel and Scotchmer (1999). The evolutionary methodology has also been utilized to study the emergence of various elements of bargaining behavior (see Young (1993) and Ellingsen (1997)). Closest to the present analysis are the study by Huck, Kirchsteiger and Oechssler (1997), explaining the "endowment effect" (the endowment effect describes the fact that people demand much more to give up an object then they are willing to spend to acquire it); and the study by Heifetz and Segev (2001) on the role of toughness in bargaining. This line of study has employed the evolutionary or cultural transmission approaches to examine how preferences are formed in different economic settings. Although psychological biases and preferences are distinct human characteristics, they are both subject to similar dynamic forces.
4
optimism. Second, if a settlement is reached, optimism provides for more favorable
settlement terms. Intuitively, optimism operates as a commitment device, leading the
litigant to appear tougher in the bargaining game. This force pushes towards more
optimism. Balancing these countervailing forces leads to the “cautious optimism” result.8
After providing a theoretical underpinning for optimism in the pretrial bargaining,
the analysis proceeds to examine which factors determine the magnitude of the optimism
bias. In particular, it is shown that the legal environment plays an important role in
shaping litigants’ perceptions. The present study focuses on the legal rules that determine
the allocation of litigation costs between the parties. It is shown that the American rule,
which lets each party bear her own litigation costs, induces a greater level of optimism,
compared to the British rule, under which the loser at trial bears also the winner’s
litigation costs.9 Intuitively, since the implications of victory at trial are greater under the
British rule10, optimism regarding the probability of victory is more powerful under this
rule. This greater impact of optimism enhances both the negative effect of optimism on
the probability of settlement and the positive effect of optimism on the terms of
settlement. Overall, this leads to a lower level of optimism under the British rule.
The higher level of optimism induced by the American rule (as compared to the
British rule) reduces the relative advantage of the American rule in fostering settlements.
8 The present analysis argues that cautious optimism provides for higher expected returns. There are
alternative explanations for the success and survival of optimism. For instance, even if optimists do poorly on average, a few lucky optimists may do better than their realist or pessimist rivals (if optimism leads players to choose lower mean, but higher variance projects or cases). Compare: Majumdar and Radner (1991) on the success and survival of risk loving behavior.
9 While the analysis focuses on the American and British rules, it can be readily extended to cover other rules for the allocation of litigation costs, such as the pro-plaintiff rule and the pro-defendant rule (for a characterization of these rules – see Shavell (1982)).
10 Under the British rule, the allocation of litigation costs as well as the judgment amount itself depend on the outcome at trial. Under the American rule, each party bears its own litigation costs, regardless of the outcome at trial.
5
These findings add a new perspective to the ongoing debate over the optimal allocation of
litigation costs. Generally, Law and Economics scholars have supported the American
rule as the more effective promoter of settlements (see, for example, Shavell (1982)).
This conventional wisdom, however, is (implicitly) based on an exogenous optimism
assumption (or simply on a no-optimism, or realism, assumption). The proposed
endogenous optimism model qualifies the conventional wisdom.
This paper demonstrates that the equilibrium level of optimism depends on the legal
environment in a systematic way. This result is indicative of a much broader theme. The
law can play an important role in determining the type and magnitude of prevailing
cognitive biases. The identification, characterization and analysis of this perception-
shaping role of legal institutions are a novelty of the present study. Behavioral law and
economics is characterized here as a two-way street, rather than the one-way
characterization implicit in the current literature. Clearly, cognitive biases affect the
operation of legal rules. But also the legal rules themselves influence the type and
magnitude of the prevailing cognitive biases. Both effects should be considered in the
evaluation of legal policy.
The remainder of the paper is organized as follows. Section 2 presents the model.
Section 3 establishes the evolutionary stability of cautious optimism in a deterministic
setting. Section 4 analyzes the effects of uncertainty on the evolutionary stable level of
optimism. Section 5 examines welfare issues and highlights the main policy implications
of the analysis. Section 6 offers concluding remarks and discusses possible extensions.
6
2. Model
2.1 Pre-Trial Bargaining
Let p represent the true probability of a judgment in favor of the plaintiff. If the
plaintiff is victorious, the defendant is ordered to pay him a sum W. The plaintiff and the
defendant incur litigation costs of pc and dc , respectively, if negotiations fail and trial
commences. Let dp ccC += represent the total costs of going to trial.
The optimism, realism or pessimism of the parties is with regard to the probability
of the plaintiff's victory at trial. Let pp xpp += represent the probability that the
plaintiff succeeds at trial, as perceived by the plaintiff. Correspondingly, let dd xpp −=
represent the probability that the plaintiff succeeds at trial, as perceived by the defendant.
Therefore, the players are considered realists if 0x = , optimists if 0x > and pessimists
if 0x < . It should be emphasized that optimism, realism and pessimism are viewed as
intrinsic characteristics of the individual, such that the individual is unaware of the bias
(if one exists).11
Under the American rule each player bears her own litigation costs regardless of the
outcome at trial. Hence, in pre-trial negotiations, the plaintiff's reservation price is
ppp cWxpR −+= )( ,12 and the defendant's reservation price is ddd cWxpR +−= )( . In
other words, the plaintiff will reject any settlement below pR , and the defendant will
11 Hence, optimism cannot be used strategically. This type of unawareness is assumed in the dynamic
extension as well (see section 2.2), where individuals play the pre-trial bargaining game again and again through time. This implies a limited bounded rationality assumption, since in our model players do not learn about their bias level.
12 The analysis assumes positive present value suits, i.e. 0>− pcpW . It is further assumed that
0)( >−+= ppp cWxpR . These assumptions focus the analysis on the effects of optimism on pre-trial
settlement negotiations, as they ensure that suit will be brought regardless of the plaintiff’s perceptional bias. It should be noted, however, that optimism may also affect the plaintiff’s decision whether to file suit in the first place. This and related effects are discussed further in section 6.1.
7
reject any settlement above dR . Therefore, a settlement range [ ]dp R,R exists if and only
if dp RR ≤ , i.e. a settlement will be reached if and only if WC
xx dp ≤+ . Under the
British rule, the loser at trial is required to cover the winner's litigation costs. Hence, the
plaintiff's reservation price is CCWxpR pp −++= ))(( , and the defendant's reservation
price is ))(( CWxpR dd +−= . A settlement range exists under the British rule if and
only if CW
Cxx dp +
≤+ . The settlement range under each of the two rules is illustrated
in figure 1.
Figure 1 exhibits the main advantage of the American Rule, as described in
traditional models. The wider settlement range induced by this rule is capable of
“absorbing” greater levels of uncertainty, and thus it guarantees more settlements.13 As
demonstrated below, this advantage of the American rule is significantly weakened when
optimism is determined endogenously (rather than assumed exogenously).
13 The American rule induces a wider settlement range under both a ‘no optimism’ (or realism)
assumption as well as under an exogenous optimism assumption.
pp cWxp −+ )( dd cWxp +− )(
Fig. 1: The settlement ranges under both rules
CCWxp p −++ ))(( ))(( CWxp d +−
American Rule
British Rule
8
If a settlement is reached, its terms will be determined by the parties' relative
bargaining power. For simplicity of exposition assume that the parties have equal
bargaining power.14 Specifically, under the American rule, if a settlement is reached, the
defendant will pay the plaintiff:
( ) ( ) ( )[ ]ddppdpA cWxpcWxpRRS +−+−+⋅=+⋅= )()(5.05.0
Similarly, under the British rule, the settlement amount will be:
( ) ( ) ( )[ ]))(())((5.05.0 CWxpCCWxpRRS dpdpB +−+−++⋅=+⋅=
Based on the aforementioned definitions and assumptions, the payoff functions in
an American regime are:
>+−
≤+=
WC
xx ,cpW
WC
xx , S
dpp
dpA
ApΠ and
( )
>++−
≤+−=
WC
xx ,cpW
WC
xx , S
dpd
dpA
AdΠ .
Similarly, in a British regime the payoff functions are:
+>+−+
+≤+
=Π
CWC
xxCCWp
CWC
xxS
dp
dpB
Bp
, )(
, and
+>++−
+≤+−
=Π
CWC
xxCWp
CWC
xxS
dp
dpB
Bd
, )(
, .
These payoff functions characterize the outcome of the pre-trial bargaining phase.
14 Alternatively, apply the Nash bargaining solution (Nash (1950)) to the pre-trial negotiations game.
Nevertheless, as will be made obvious shortly, the main results of this paper do not hinge upon the assumption of equal bargaining power (nor on the alternative assumption of a Nash solution).
9
2.2 Matching, Fitness and Evolutionary Dynamics
The analysis assumes a single population of potential plaintiffs and defendants.
Each period the population is randomly divided into two groups, such that half the
population are plaintiffs and half the population are defendants. Each member of the
plaintiff group is randomly matched with a member of the defendant group.15
Let ix represent the level of optimism (or pessimism) of individual i. The level of
optimism is independent of the role, plaintiff or defendant, which is assigned to player i at
the beginning of each period.
The distribution of optimists, realists and pessimists in the population is represented
by the probability density function (pdf) – )(xg , and the corresponding commulative
distribution function (cdf) – )(xG .
The )(xg function determines the probability that player i will meet a player j with
an jx level of optimism. The probability that a player i will be assigned the role of
plaintiff and meet a defendant j is equal to the probability that i will be assigned the role
of defendant and meet a plaintiff j. Therefore, the a-priori expected payoff to a player i
which meets a player j in an American regime is:
15 A different model may consist of two distinct populations - a plaintiff population and a defendant
population. This mo del better describes a reality, in which a certain group of people is more likely to sue, while a different group is more likely to be on the defending side. The model studied in this paper better describes a reality, in which the entire population has equal a-priori probabilities of becoming plaintiffs or defendants. As in the single-population model (see section 3), the two-population model also yields the basic cautious optimism result. Nevertheless, the analysis of the two-population model employs a slightly different methodology.
10
[ ]
( ) ( )[ ]
( ) ( )[ ]
( ) ( )[ ]dipjAd
djpiAp
jidp
jiAd
Ap
jiAi
cWxpcWxpS
cWxpcWxpSwhere
WC
xxcpWcpW
WC
xxSSxxm
+−+−+=
+−+−+=
>++−−
≤+−=
)()(21
)()(21
, 21
, 21
),(
or -
[ ]
( )
>+−
≤+−=
WC
xxC
WC
xxWxxxxm
ji
jiji
jiAi
, 21
, 21
),( )1(
The expected payoff function in a British regime is -
[ ]( )
( )
+>+−
+≤++−
=
CWC
xxC
CWC
xxCWxxxxm
ji
jiji
jiBi
, 21
, 21
),( )2(
The expected fitness of player i is defined as -
∫=x
iiii dxxgxxmxf )(),()( )3(
A description of the evolutionary dynamics completes the formulation of the model.
As specified earlier, at period t each player i is randomly assigned a role of plaintiff or
defendant and is then matched with a player j of the opposite role according to the current
distribution of players in the population, )(xg t . Using equation (3) the expected fitness
of every player i can be calculated, yielding the fitness function, )(xf t . The fitness
11
function, together with the current distribution, determines the distribution of optimism in
the following period, via a payoff/fitness-monotonic dynamic algorithm.16
Sections 3 proceeds to study the evolutionary stable distributions of optimism and
to prove that cautious optimism will prevail.
3. The Evolution of Optimism
3.1 Optimism - Pros and Cons
The question whether optimism will eventually dominate the population, and if so
at what level, depends upon the relative strength of the following two effects, which
determine the impact of a variation in the level of optimism on a player's payoffs /
fitness:17
Effect 1: A higher level of optimism entails fewer settlements (and consequently more
costly trials).
Take the American rule, for instance. Substituting expression (1) into
expression (3), results in the following:
( ) ( )[ ]
−+−
−−= ∫
− ixWC
xiii dxxgxxWCx
WC
Gxfmin
)(121
)( (4) i
16 The results presented below hold for any payoff monotonic dynamic process. See Weibull (1995) for
examples of commonly used payoff-monotonic dynamic algorithms. The numeric calculations (see appendix B) employ the discrete-time (continuous strategy) version of the replicator dynamics:
{ }.)(min)()(ˆ , )()(ˆ)()(ˆ
)(1 xfxfxfwheredxxgxf
xgxfxg txtt
tt
ttt −≡
⋅
⋅=
∫+
17 These two effects may be viewed as stylized facts, which are captured, in reduced form, by the present
model.
12
A player i with a ix level of optimism will reach a settlement with any player j
with a level of optimism ij xWC
x −≤ . Hence, more optimistic individuals will
reach a settlement with a smaller range of players. However, it should be noted,
that a settlement is not necessarily preferable to trial. A pessimistic player may
reach a settlement, which is worse than the expected outcome at trial, including
litigation costs.
Effect 2: A higher level of optimism guarantees more favorable settlements.
A player's optimism shifts the settlement range in her favor (optimistic plaintiffs
shift the settlement range towards higher settlements, and optimistic defendants
shift the settlement range towards lower settlements).18 This effect is captured
by the integral term of equation (4) above.
At the evolutionary equilibrium effect 1 and effect 2 balance out on the margin.19
3.2 The Disappearance of Pessimism
Pessimistic players will not survive the evolutionary selection process. This result is
summarized in the following proposition.
18 The plaintiff’s reservation price is pcWxp −+ )( , therefore a more optimistic plaintiff will shift the
settlement range towards higher settlements. The defendant’s reservation price is dcWxp +− )( ,
therefore a more optimistic defendant will shift the settlement range towards lower settlements. 19 Similar effects are noted in Bottom and Paese (1999) and in Cooter, Marks and Mnookin (1982). See
also Huck, Kirchsteiger and Oechssler (1997), who identify two competing effects with regard to the "endowment effect". On the one hand, a positive "endowment effect" distorts the player's substitution rate between the good she is endowed with and the good she needs to acquire. On the other hand, the "endowment effect" improves the player's bargaining position.
13
Proposition 1: Under both the American and British rules, pessimists, i.e. players with
0<x , will disappear, namely will not survive the evolutionary selection process –
0)(lim 0t
=<∀∞→
xgx t
Remark: The intuition for this result, which is proved in appendix A, relies on the two
effects of optimism, which were identified in section 3.1. Effect 2 encourages pessimistic
players to become more optimistic. Therefore, it remains to be shown that effect 1 also
supports more optimism. Effect 1 states that increased optimism leads to more trials.
However, a player with 0<x prefers going to trial to an out of court settlement, i.e. the
settlement is less preferable than the expected judgment including litigation costs. The
reason is that the extra settlements reached by pessimists are settlements with very
optimistic rivals (a realist, and even a cautious optimist, will settle with a mildly
optimistic rival). These extremely optimistic rivals will drive such a hard bargain that the
pessimist is better off going to trial.
3.3 Evolution Towards Cautious Optimism
Proposition 1 has shown, that pessimism will be abolished in the evolutionary
process. It remains to be seen whether realism is evolutionary stable or perhaps optimism
is evolutionary stable, and if so what level of optimism. Proposition 2 begins by
demonstrating that evolutionary forces eliminate excessively optimistic individuals.
14
Proposition 2: (i) Under the American rule, excessively optimistic players, with WC
x > ,
will disappear, namely will not survive the evolutionary selection process -
0)(lim t
=>∀∞→
xgWC
x t
(ii) Under the British rule, excessively optimistic players, with CW
Cx
+> , will
disappear, namely will not survive the evolutionary selection process -
0)(lim t
=+
>∀∞→
xgCW
Cx t
Remark: The intuition for this result, which is proved in appendix A, is as follows. Using
the two effects of optimism from section 3.1, note that optimists will generally enjoy the
benefits of effect 2, which promises more favorable settlements. However, after the
disappearance of pessimists, as shown in proposition 1, players with WC
x > will never
settle under the American rule, and players with CW
Cx
+> will never settle under the
British rule. Therefore, after the disappearance of pessimism, effect 2 becomes moot,
leaving only effect 1, which supports a decrease in the level of optimism.
Proposition 2, combined with proposition 1, narrows down the range of possible
bias levels significantly -
Proposition 3: Under both the American and British rules, in the evolutionary
equilibrium, all players will be either realistic or cautiously optimistic.
15
Under the American rule -
0)(lim ,0t
=
∉∀
∞→xg
WC
x t
Under the British rule -
0)(lim ,0t
=
+∉∀
∞→xg
CWC
x t
Remark: Proposition 3 is derived directly from combining the results of proposition 1 and
proposition 2.
Proposition 3 takes a large step towards one of the paper's main results, namely the
evolutionary tendency towards "cautious optimism". All that remains to be shown is that
the population will not converge to the lower boundary of realism. However, the analysis
leads to a much stronger result. Not only does it prove the non-stability of realism, it also
pin-points the level of optimism in equilibrium, under the two legal regimes. This result
is derived using the following two lemmas.
Lemma 1: (i) Under the American rule, the degenerate distribution g(x), in which the
entire population shares the unique level of optimism WC
x̂2
= is evolutionary stable.
(ii) Under the British rule, the degenerate distribution g(x), in which the entire population
shares the unique level of optimism )(2
ˆCW
Cx
+= is evolutionary stable.
16
Remark: The intuition for this result, which is proved in appendix A, is as follows.
Focusing on part (i) of the lemma (part (ii) follows a similar intuition), consider a
homogeneous WC
x̂2
= -population. Can any “mutant” with a different level of optimism
successfully invade this population, i.e. can any such mutant earn a payoff higher than the
payoff earned by members of the population? The answer is ‘no’. Members of the
population, when matched-up with one another, always settle and earn a zero payoff
( 0)ˆ,ˆ( =xxmi ). A mutant with WC
xm 2< , when matched-up with a member of the
population, will always settle as well; and thus by effect 2 will be hurt by her lower level
of optimism. A mutant WC
xm 2> , when matched-up with a member of the population,
will never settle, and thus will incur substantial litigation costs.
Lemma 2: (i) Under the American rule, any degenerate distribution g(x), other than the
distribution in which the entire population shares the unique level of optimism WC
2, is
evolutionary unstable.
(ii) Under the British rule, any degenerate distribution g(x), other than the distribution in
which the entire population shares the unique level of optimism )(2 CW
C+
, is
evolutionary unstable.
Remark: The intuition for this result, which is proved in appendix A, is as follows.
Focusing on part (i) of the lemma (part (ii) follows a similar intuition), consider in turn a
17
homogeneous WC
x̂2
> -population and a homogeneous WC
x2
ˆ < -population. Starting
with the homogeneous WC
x̂2
> -population, a small group of mutants with WC
xm 2= can
successfully invade this population, i.e. such mutants will earn a payoff higher than the
payoff earned by members of the population. Members of the population will never settle
– neither when matched-up with one another nor when matched up with the mutants. The
mutants also will not settle, when matched-up with a member of the population, but they
will settle, when matched-up with their own kind. Therefore, the WC
2-mutants save on
litigation costs, and hence earn higher payoffs.
Moving on to the homogeneous WC
x2
ˆ < -population, a mutant with WC
xm 2= can
successfully invade this population as well. Members of the population, when matched-
up with one another, always settle and earn a zero payoff ( 0)ˆ,ˆ( =xxmi ). A mutant with
WC
xm 2= , when matched-up with a member of the population, will likewise settle, but
according to effect 2 will enjoy more favorable settlements. Therefore, the WC
2-mutants
earn higher payoffs.
A population of realists, namely a homogenous population where everyone shares
the 0ˆ =x level of optimism, is a special case within the category of homogeneous
WC
x2
ˆ < -populations. Lemma 2 precludes the possibility of a population composed solely
of realists.
18
Combining lemma 1 and lemma 2 results in the following proposition –
Proposition 4: (i) Under the American rule, the unique evolutionary stable homogenous
population shares the WC
x̂2
= level of optimism.
(ii) Under the British rule, the unique evolutionary stable homogenous population shares
the )(2
ˆCW
Cx
+= level of optimism.
Remark: Proposition 4 is derived directly from combining the results of lemma 1 and
lemma 2.
Proposition 4 summarizes the influence of the legal rule on the equilibrium level of
optimism. The American rule is shown to induce a higher level of optimism, as compared
to the British rule. Intuitively, since the implications of victory at trial are greater under
the British rule20, optimism regarding the probability of victory is more powerful under
this rule. This greater impact of optimism enhances both the negative effect of optimism
on the probability of settlement and the positive effect of optimism on the terms of
settlement. In essence, a lower level of optimism under the British rule has the same
effect as a higher level of optimism under the American rule.
20 Under the British rule, the allocation of litigation costs as well as the judgment amount itself depend on
the outcome at trial. Under the American rule, each party bears its own litigation costs, regardless of the outcome at trial.
19
The remainder of the paper focuses on homogenous populations of the kind
characterized in proposition 4.21 Substituting the stable levels of optimism from
proposition 4 into the definition of the settlement range (see section 2.1), leads to the
following corollary –
Corollary: Under both the American and British rules the evolutionary equilibrium
induces a degenerate settlement range, i.e. dp RR = .
21 Heterogeneous stable populations also exist in the present model. For instance, the following 2-type
population is evolutionary stable: ( )
⋅=−=
∈ 112121
2 , ,
2,0 ,, x
CW
pxWC
xw
Cxpxx ,
where 1x and 2x are the optimism levels of the two types and p is the proportion of the first type (with
optimism 1x ) in the population. The average bias level in these populations may also be characterized as cautious optimism. Nevertheless, we choose to focus on the unique stable homogenous population for the following reasons: First, numeric calculations demonstrate that the stable homogenous population has a particularly large basin of attraction. Specifically, if the initial distribution of optimism in the population is not “too concentrated” outside the ranges defined in proposition 3 ( ],0[ WC for the American rule and )](,0[ CWC + for the British rule), the dynamic process will converge to the stable homogenous population (see appendix B). Second, since no a-priori heterogeneity was assumed (except for possibly different initial levels of optimism) a resulting stable homogenous population seems intuitively appealing. Moreover, professional norms within the legal profession foster such
homogeneity. Third, the WC
x2
ˆ = (or )(2
ˆCW
Cx
+= under the British rule) result emerges also via a
completely different approach to the pre-trial bargaining game. Assume that litigants observe lawyers’ optimism (directly or through its correlation with their expected earnings), and thus may choose a lawyer, and consequently a level of optimism, as if it were a strategy in a bargaining game (see section
6.7 below). In such a 2-player game between a plaintiff and a defendant, WC
xx dp 2== is the unique
symmetric Nash equilibrium, and it is also supported by the Nash bargaining solution (Nash (1950)). Now, consider a 2-player game, in which each player has an equal chance of being either a plaintiff or a
defendant. The Nash solution supports the WC
x2
= “strategy” for both players in this game as well.
Finally, consider a N-player game, in which each player has an equal chance of being “assigned” to either the plaintiff group or the defendant group (as described in the text). Here too a natural extension
of the Nash solution supports the common WC
x2
= “strategy”.
20
Remark: This result follows immediately from the definitions of pR and dR (see section
2.1).
Recall from section 2.1 that under the realism assumption or under the alternative
assumption of exogenous optimism (or pessimism), the settlement range under the
American rule is wider than the settlement range under the British rule. For this reason
the American rule is commonly considered to be a better promoter of settlements. The
preceding corollary qualifies this advantage of the American rule when optimism is
allowed to evolve endogenously according to the prevailing legal regime.
It should be noted, however, that in the present deterministic setting all pre-trial
negotiations, under both rules, result in settlement, even when the (endogenous) optimism
bias is accounted for. In other words, endogenous optimism by itself does not prevent
settlements. But, optimism rarely operates in the sterile deterministic setting postulated in
the above model. Therefore, a quantitative evaluation of the effect of endogenous
optimism on the probability of settlement requires a synthesis between the endogenous
optimism model and the standard models of uncertainty and asymmetric information,
which have been commonly used to explain negotiation failures.
A comprehensive synthesis between endogenous optimism and traditional
impediments to bargaining, like uncertainty and asymmetric information, is beyond the
scope of the present study. Nevertheless, the following section offers an illustrative
example that demonstrates how endogenous optimism affects standard results in a
combined model of uncertainty and optimism.
21
4. Optimism and Uncertainty – An Example
4.1 Adding Uncertainty
Section 4 adds a simple form of uncertainty to the preceding deterministic model,
and studies the endogenous evolution of optimism in the resulting stochastic framework.
The stochastic model will then be used to quantify the effects of endogenous optimism on
the probability of settlement (see section 5). Naturally, uncertainty may enter into the
model in a variety of ways. Litigants may be uncertain about the probability of the
plaintiff’s victory (p), about the magnitude of the judgment (W), about their own
litigation costs, about their opponent’s litigation costs, and so on. Abstracting from all but
one type of uncertainty, it is assumed that the defendant observes the probability, p, with
a noise. Namely, ε+= ppd , where ε is a symmetrically distributed noise element. Note
that the plaintiff observes p correctly, but cannot convey this information to the
defendant.22
The noise element affects the choice between settlement and trial through its
influence on the defendant’s reservation price, dR . When these effects are accounted for,
the a-priori expected payoff to a player i, who meets a player j in an American regime
(expression (1)), becomes:
22 Certainly, in many bargaining situations this assumption is unrealistic, at least at its extreme. Section 4,
however, adopts this assumption for the sake of expositional clarity and analytical convenience. The main results carry over to more complex (and more realistic) assumptions regarding the nature of the uncertainty and the structure of the information. Specifically, results similar to the ones presented below have been derived numerically in a model where both plaintiff and defendant observe a “noisy” p (in such a model our assumption regarding transfer of information between the parties becomes more plausible). The numeric simulations apply a Monte Carlo algorithm: The algorithm randomly selects a noise level for each player (from a given noise distribution), and then lets each plaintiff-defendant pair play out the litigation-settlement game with their respective noisy perceptions of p. See appendix B for further details.
22
[ ] ( )
( ) ( )
( ) ( )[ ]
−+⋅+−−−=
−+<−
−+≥−= ∫
WC
xxHCWxxWxx
dh
WC
xxC
WC
xxWxxxxm
jijiji
ji
jiji
jiAi
21
21
)( ,
21
, 21
),( )5( εεε
ε
where h(.) is the symmetric pdf of the noise element, and H(.) is the corresponding cdf.23
In a British regime the “noisy” parallel of expression (2) is:
[ ] ( )
( ) ( )
( ) ( )[ ]
+−+⋅++−−+−=
+−+<−
+−+≥+−
= ∫
CWC
xxHCCWxxCWxx
dh
CWC
xxC
CWC
xxCWxxxxm
jijiji
ji
jiji
jiBi
)(21
)(21
)( ,
21
, )(21
),( )6( εεε
ε
4.2 Evolution of Optimism under Uncertainty
The analysis proceeds to examine how uncertainty affects the evolutionary stable
level of optimism. Taking the WC
x̂2
= result of section 3 as a benchmark, it is first
shown that this result no longer holds when uncertainty is added to the model. The
evolutionary stable level of optimism depends on the level of uncertainty. The
relationship between the level of uncertainty and the equilibrium level of optimism is
then derived. The subsequent analysis concentrates on the American rule.24 The analysis
of the British regime is virtually identical.
Differentiating the payoff function (expression (5)) with respect to ix yields – 23 The formulation of the payoff function in expression (5) is based on another simplifying assumption
regarding the noise element. It is assumed that the uncertainty is relationship specific in the sense that a single realization of ε occurs for each i-j pair, regardless of the role assignment (of plaintiff and defendant) between i and j.
24 Hence, the superscript A is omitted from the payoff functions for notational convenience.
23
[ ]
−+⋅+−⋅−
−+⋅⋅−⋅=
WC
xxhCWxxWC
xxHWWdx
xxdmjijiji
i
jii )(21
21
21),(
)7(
Focusing on homogenous populations, let xx j ˆ= represent the level of optimism in the
population, and let mi xx = represent the level of optimism of a possible mutant. If the
homogenous x̂ -population is evolutionary stable, then xxm ˆ= maximizes the payoff
function, and the derivative m
mm
dxxxxdm )ˆ,ˆ( =
equals zero.
Starting from the homogenous WC
x̂2
= -population, and substitute WC
x̂xm 2==
the derivative function becomes:
( ) ( )[ ]
⋅⋅−=⋅−⋅−=
==)0(21
41
0021)ˆ,(
)8( 22 hWC
WhCHWWdx
xxdm
m
WC
WC
mm
Hence, generally, a mutant would be able to successfully invade a homogenous WC
x̂2
= -
population. Note, that the key element of expression (8) is )0(h , which is determined by
the variance of the noise element (recall that we assume a symmetrically distributed
noise). Specifically, )0(h is a decreasing function of the variance of the noise
distribution. Therefore, when the level of uncertainty is low, mutants with WC
xm 2< will
successfully invade the population, and when the level of uncertainty is high, mutants
with WC
xm 2> will successfully invade the population. Note, however, that if )0(h is
continuous in the variance of the noise distribution, there exists an intermediate level of
uncertainty for which the homogenous WC
x̂2
= -population is evolutionary stable.
24
Departing from the WC
x̂2
= benchmark, the next step is to derive the stable level of
optimism as a function of the noise level, as captured by the standard deviation of the
noise distribution εσ . Naturally, the specific optimism levels depend on the noise
distribution. The analytical solution presented below assumes a triangular noise
distribution. However, the general characteristics of the )(ˆ εσx function are common to
all symmetric noise distributions. Substituting the triangular pdf and cdf into expression
(7), and solving for the stable level of optimism, we obtain -
⋅>+−
⋅≤
++−
=
W
CWC
WC
WC
x A
5.1 , 5.1
2
5.1 ,
235.1
)(ˆ
21
22
εε
εεε
ε
σσ
σσσσ
Using parallel reasoning, the )(ˆ εσBx function can be derived for the British rule
(with a triangular noise distribution):
( )
( )
+⋅>+
+−
+⋅≤
+
++−=
CW
CCW
C
CWC
CWC
x B
5.1 , 5.1
)(2
5.1 ,
)(235.1
)(ˆ
21
22
εε
εεε
ε
σσ
σσσσ
Figure 2 compares the stable levels of optimism under the two legal rules for
different noise levels.
25
The intuition for the results depicted in figure 2 is as follows. When uncertainty is
added into the model lower optimism opens the degenerate settlement range (which was
derived in section 3 in a deterministic setting) and allows it to “absorb” the uncertainty.
This raises the number of settlements and increases the monetary payoffs. But lowering
the level of optimism also entails a cost – the cost of less favorable settlements. For low
levels of uncertainty the former effect dominates and the stable level of optimism
decreases. However, for higher levels of uncertainty the cost of ensuring a settlement
becomes too high. Moreover, as the level of uncertainty increases players with higher and
Fig. 2: Stable levels of optimism under the American and British rules (parameter values: W = 500 , C = 100)
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350.05
0.1
0.15
0.2
0.25
0.3
std of noise element
stab
le le
vel o
f op
timis
m
American RuleBritish Rule
26
higher levels of optimism, who could never settle in the deterministic setting,
occasionally reach settlements. This explains the rising graphs in the high-noise range.
After analyzing the evolution of optimism in both a deterministic and a stochastic
setting, it remains to examine the welfare and policy implications of the endogenous
optimism phenomenon.
5. Welfare and Policy Implications
Sections 3 and 4 have demonstrated that evolutionary forces drive the parties
towards cautious optimism with regard to the probability of prevailing at trial. The
optimism phenomenon, which has been empirically observed, and has now also received
a theoretical explanation, has important welfare and policy implications.
First and foremost, optimism is detrimental to settlement. The dynamic model
analyzed in section 3 has shown that evolution favors a moderate level of optimism.
Section 4 demonstrated, how the evolutionary stable level of optimism varies with
respect to the degree of uncertainty. Given the derived stable level of optimism and the
noise distribution, the probability of settlement can be calculated. In the present context,
the sole welfare consideration is the maximization of the settlement probability, leading
to a minimization of litigation costs. Hence, the models employed in sections 3 and 4 can
be readily extended to include welfare analysis.
Using the results of the previous sections, the relationship between the level of
uncertainty and the probability of settlement can be derived. Uncertainty and optimism
are both detrimental to settlement. For low levels of uncertainty, an increase in the degree
of uncertainty lowers the stable level of optimism (as shown in figure 2). Hence, the two
27
effects work in opposite directions. The direct effect of uncertainty, however, is
dominant, and thus welfare drops when uncertainty rises, even for low levels of
uncertainty. For high levels of uncertainty, an increase in the degree of uncertainty raises
the stable level of optimism. Hence, welfare clearly drops.
That uncertainty decreases welfare is not surprising. The interesting comparison is
between the probability of settlement under the standard assumption of realism, as
opposed to the probability of settlement in our endogenous optimism model. Figure 3
depicts the probability of settlement, under the American rule, as a function of the level
of uncertainty – first, when realism is exogenously imposed, and second when
endogenous optimism is considered (the quantitative results were derived using the
framework developed in section 4).
Fig. 3: The probability of settlement as a function of the noise level (parameter values: W = 500 , C = 100)
0 0.1 0.2 0.3 0.4 0.50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
std of noise element
prob
abili
ty o
f se
ttle
men
t
endogenous optimismrealism
28
Low levels of uncertainty hardly affect the standard models, as they are completely
“absorbed” by a wide settlement range. For this reason realism models resort to
asymmetric information in order to explain negotiation failure. Given parties’ optimism,
on the other hand, the shrinkage of the settlement range renders the settlement
opportunities volatile, in a way that even the slightest symmetric noise may prevent
settlement. For higher levels of uncertainty, negotia tions may fail even under the
assumption of realism. Still, the probability that a settlement will not be reached is
significantly higher, when optimism is accounted for. This result supports an increased
concern regarding the vagueness and uncertainty that surround the legal process. More
precise procedural rules, narrower judicial discretion and more efficient discovery rules -
are all policy instruments supported by the present analysis.
A second policy implication, albeit a more tentative one, concerns cost-shifting
rules. Classical analysis suggests that the American rule is preferable to the British rule,
since it induces a larger settlement range (see, for example, Shavell (1982) and Posner
(1998)). The evolutionary analysis refines this traditional understanding. The endogenous
optimism model proves, that in a deterministic setting the settlement range will disappear
under both rules, and the different rules only affect the level of optimism at equilibrium.
When a symmetric noise is introduced into the model, the size of the settlement range
depends on the degree of uncertainty, under both regimes. However, the functional
relationship between the degree of uncertainty and the evolutionary stable level of
optimism, which determines the size of the settlement range, varies from one legal
regime to the other (see section 4). Therefore, a comparison, in terms of welfare, between
the American rule and the British rule, is not one-dimensional. Such a comparison
29
depends on the specific values of the model's parameters and especially on the degree of
uncertainty.
Consider the relationship between the probability of settlement and the level of
uncertainty under the two regimes, as derived using the model developed in section 4
(with a triangular noise distribution). The results are presented in figure 4.
As figure 4 demonstrates, the probability of settlement is higher under the American
rule. This result is consistent with traditional models that impose exogenous realism or an
exogenous level of optimism. It is interesting, however, to compare the advantage of the
American rule under an exogenous optimism assumption and under endogenous
optimism. The following example is illustrative. Assume that the stakes at trial are W =
Fig. 4: The probability of settlement under the two legal rules (parameter values: W = 500, C = 100)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.70
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
std of noise element
prob
abili
ty o
f set
tlem
ent
American RuleBritish Rule
30
500 and that the parties’ total litigation costs are C = 100. Also, assume that the standard
deviation of the noise element is 0.05 (assuming a triangular noise distribution). First,
consider endogenous optimism. Under the American rule the level of optimism will be
x = 0.071 and the probability of settlement will be 86%. Under the British rule the level
of optimism will be x = 0.059 and the probability of settlement will be 82%. However, if
the level of optimism is determined exogenously to be, for instance, x = 0.071, the
probability of settlement under the American rule is still 86%, but the probability of
settlement under the British rule is only 68%. This simple numeric example demonstrates
that endogenous optimism may be a significant phenomenon. It also suggests that
assuming an exogenous level of optimism, as is common in the literature, may turn out to
be quite misleading.
6. Concluding Remarks and Proposed Extensions
6.1 Optimism in the Decision to Bring Suit
Focusing on settlement negotiations, rather than on the decision to file suit, the
preceding analysis has assumed that all claims are perceived to have (and indeed have) a
positive net present value, regardless of the plaintiff’s level of optimism. However,
optimism may also affect the decision to bring suit. In particular, an optimistic plaintiff
may decide to file a negative present value (NPV) suit. This NPV suit may settle, thus
providing another advantage of optimism. On the other hand, the NPV suit might lead to
a trial that by definition will impose a net loss on the plaintiff; another disadvantage of
optimism.
31
A complete analysis of the “decision to bring suit” extension is beyond the scope of
the present study. Still, the preceding observations suggest that the general “cautious
optimism” result is robust to such an extension. Moreover, the general theme that legal
institutions affect the magnitude of prevailing cognitive biases clearly carries over to the
“decision to bring suit” extension.
6.2 Other Self-Serving Biases and Toughness in Bargaining
The preceding analysis focused on the optimism bias. Optimism, however, is not
the only cognitive bias that may influence the pretrial bargaining process. In particular,
the psychology literature has identified a class of self-serving biases, which have been
shown to exist in the pretrial environment (see Lowenstein et al. (1993)). The model
studied is this paper can be readily adjusted to analyze these self-serving biases.
Related to the self-serving biases is the notion of toughness in bargaining, namely
that “individuals enter a tough state of mind when they have to make a stand vis-à-vis
somebody else”. 25 The proposed model can also be adjusted to study the evolutionary
stable degree of toughness.
6.3 Information Acquisition
The endogenous optimism model, as well as the extensions considered above,
deviate from the neo-classical view of the rational decision-maker. However, the analysis
also suggests a meaningful interpretation within the neo-classical framework. Assume
that litigants do not “suffer” from genuine optimism or pessimism. Instead, litigants or
25 See Heifetz and Segev (2001). These authors study an evolutionary model of toughness in general
bargaining contexts. (The working paper by Heifetz and Segev was issued a year after the working paper that preceded the current paper, Bar-Gill (2000).)
32
lawyers form rational (unbiased) estimates of their chances at trial based on the
information available to them. The question then becomes what information sources are
available to these litigants and lawyers and what methods of gathering information they
employ. If the lawyer has control over her sources of information and can choose among
various methods of information acquisition, then the analysis of the present paper
becomes relevant once again. The lawyer’s choice of sources and methods for acquiring
information parallels the quasi-evolutionary selection of a particular bias level. In
analogy, the present model would predict a tendency of lawyers to opt for information
sources and for information gathering methods that produce biased self-serving estimates
of their chances at trial.
6.4 The “Zealous Advocacy” Norm
The results presented in this paper, together with the extensions considered above,
suggest a certain level of optimism, toughness and the production of intentionally skewed
information in pretrial negotiations. Moreover, the analysis predicts greater levels of
optimism, toughness and skewed information under the American rule, as compared to
the British rule. This prediction is in line with the prominence of the “zealous advocacy”
norm in the United States.26 In contrast, the ethical rules of conduct in England
substantially qualify the reach of the zealous advocacy norm (see Osiel (1990) and Kagan
(1994)).
26 See Canon 7 of the 1969 Model Code of Professional Responsibility: “A Lawyer Should Represent A
Client Zealously Within the Bounds of Law.”
33
6.5 Lawyers and Other Repeat Players
The proposed model assumes that a single group of players repeatedly interact with
each other in litigation and settlement games. This simplifying assumption seems
especially appealing for lawyers and other repeat players (e.g. insurance companies).27
The multiple interactions within this group guarantee a relatively quick convergence to
the evolutionary equilibrium. Hence, if indeed most civil cases are managed by attorneys,
the endogenous optimism effect may be quite significant. In addition, the model suggests
a new explanation to the commonly observed advantage of repeat players in litigation.
Presumably, these repeat players will be quicker to adopt an efficient bias level and
consequently will perform better than one-time litigants. Still, these conjectures need to
be verified and possibly refined within an explicit model of both repeat and one-shot
players.
6.6 Optimism vs. Reputation
The evolutionary approach adopted in the present study requires a dynamic setting
with repeated interactions. The repeated interactions assumption, in turn, suggests the
possibility that the litigants and lawyers develop a reputation for toughness in bargaining.
In the present context, optimism and reputation may be viewed as substitutes. They both
serve as commitment devices in the negotiation game. This paper’s analysis sheds light
on the type of reputation that a lawyer would like to develop. On the one hand, a
reputation for “soft” bargaining is undesirable. On the other hand, a reputation for
27 Note that the move from litigants’ optimism to lawyers’ optimism is not a trivial one. In particular, the
lawyer’s fee structure and the market for legal services must be considered in the formulation of the lawyer’s payoff function. However, adopting the simplifying assumption of a competitive market for lawyers, in which only lawyers who maximize their clients’ profits survive, the present analysis may be applied directly to lawyers’ optimism. See also section 6.7 below.
34
stubbornness or excessive aggression at the bargaining table should also prove to be
counterproductive. Thus, in analogy to the endogenous optimism model, a reputation for
some limited degree of toughness in bargaining would seem to be a popular trait.
The relative importance of the optimism effect and the reputation effect is context
dependent. For reputation to develop information regarding the bargaining behavior of
one lawyer in a specific case must be transmitted to other lawyers. While clearly
plausible in certain close-knit communities, the informational requirements for the
development of substantial reputation effects are often not satisfied. Consider a more-or-
less competitive market for litigation services, where interactions between lawyers are
practically anonymous. In such a market, it will be more difficult for an individual lawyer
to develop the desired reputation. Anonymity, however, need not hinder the operation of
market selection forces of the type invoked in the endogenous optimism model. A
pessimist, or excessively optimistic lawyer will lose money until she is driven out of the
anonymous market. In addition, the transmission of optimism via imitation often imposes
less stringent informational requirements, relative to the informational prerequisites for
the development of accurate reputation. Finally, however, optimism and reputation may
work in concert. Market selection forces will work best when the success (failure) of a
lawyer with the “correct” (or “incorrect”) level of optimism or toughness is made widely
known through reputation mechanisms.
6.7 Evolution versus Strategic Delegation
The basic model did not distinguish between the litigants themselves and their
lawyers. The lawyer–client agency problem is, of course, well recognized, and has been
35
the subject of extensive research. This agency relationship suggests several extensions to
the preceding analysis. First, consider the following alternative selection process.
Lawyers can develop a reputation for specific levels of optimism or toughness. Clients
can then simply choose a level of optimism by choosing a lawyer with that level of
optimism. 28 It can be shown that in the game between two litigants, who choose lawyers
and optimism levels as if they were mere strategies, “cautious optimism” will prevail at
equilibrium.29, 30
A second extension abstracts from reputational considerations and focuses on
contractually designed incentive schemes. The question here is how to design the lawyer
– client contract, so as to give the lawyer optimal incentives in the strategic interaction
with rival lawyers. In particular, can the lawyer’s fee structure be designed to induce the
lawyer to negotiate as if she were cautiously optimistic? The strategic delegation problem
is well known in the industrial organization literature (see Fershtman and Judd (1987);
see also Kyle and Wang (1997)).31 In the litigation and settlement context, Bebchuk and
28 In the present interpretation, reputation operates to convey information from lawyers to clients, and not
between potential bargaining rivals (compare section 6.6 above). This interpretation relates to an argument developed by Gilson and Mnookin (1994) that clients choose lawyers with a reputation for cooperation.
29 In a simultaneous-move game, where the plaintiff chooses a lawyer with a level of optimism px and
the defendant chooses a lawyer with a level of optimism dx , the set of Nash equilibria
( ) [ ]{ }WCxxWCxxxx dpdpdp =+∈ , ,0, ,
restricts attention to “cautious optimism”, as defined in proposition 3 (and WC
xx dp 2== is the
only symmetric Nash equilibrium). See also note 20 above. 30 If the client’s perceptions, rather than the lawyer’s perceptions, are determinative in settlement
negotiations, then the analysis suggests a new role for lawyers - to misrepresent (to the client) the true chances of prevailing at trial. Optimistic misrepresentations may increase the client’s expected returns. Of course, such misrepresentation violates lawyers’ ethical norms. Indeed, from a social perspective that values a higher probability of settlement, but is indifferent to the distribution of value between litigants, there is good reason to enforce these ethical norms.
31 This paper is not the first to examine the relationship between evolution and strategic delegation. See Guth and Dufwenberg (1999).
36
Guzman (1996) compare the incentive effects of a contingent fee contract and an hourly
fee contract. Still, a comprehensive analysis of the optimal lawyer – client contract is still
missing, and must be left for future research.
6.8 The Perception-Shaping Role of Law and Two-Way Behavioral Law and Economics
A main purpose of the present analysis was to lay a theoretical foundation for the
commonly observed optimism in legal settings. Such a theoretical foundation was
missing in the large body of literature, which studies the effects of optimism without any
thorough account of its origin. Moreover, the present analysis demonstrated how the legal
regime determines the evolutionary stable level of optimism. The effects of legal rules on
the type and magnitude of prevailing cognitive biases is a novelty of the present study
(see also Bar-Gill (2002)). Behavioral law and economics is characterized here as a two-
way street, rather than the one-way characterization implicit in the current literature.
Clearly, cognitive biases affect the operation of legal rules. But also legal rules and legal
institutions influence the type and magnitude of the prevailing cognitive biases. Both
effects should be considered in the evaluation of legal policy.
6.9 From Endogenous Optimism to Endogenous Preferences
The present study has demonstrated the potential effects of legal institutions on the
type and magnitude of prevailing cognitive biases. The evolutionary influence of the law
may carry an even greater weight when the focus of the analysis is shifted from cognitive
biases to preferences. The evolutionary methodology employed in this paper may be
readily extended to analyze the evolution of preferences under different legal regimes.
37
The analysis suggests that different legal rules may lead to different stable profiles of
preferences. Consider, for example, the evolution of a preference for fairness that affects
the terms of settlement deemed acceptable by a litigant.32 The intensity of such
preferences may depend on the legal environment.
This uncharted effect of the legal system imposes a serious philosophical caveat on
any proposed legal policy. Even if the proposed policy is optimal given the current
preference profile, will it still be optimal after the new policy induces a new preference
profile (see Bar-Gill and Fershtman (2000)). These perplexing questions are of the utmost
importance. Nevertheless, they must be left open for future research.
32 Farmer and Pecorino (2000) study the effects of a preference for fairness in the context of pretrial
settlement negotiations. These authors, however, consider only the possibility of an exogenous pre-determined preference for fairness.
38
Appendix A: Proofs
Appendix A collects the proofs of propositions 1, proposition 2, lemma 1 and
lemma 2.
Proof of Proposition 1: The proof presented below is for the American rule. The proof for
the British rule follows a similar logic.33 The proof consists of two stages. Stage 1 shows
that pessimism is weakly dominated by realism.34 Stage 2 utilizes the result of stage 1 to
establish the disappearance of pessimism through the evolutionary selection process.
Stage 1: Stage 1 proves that 0=ix weakly dominates 0<'ix . Substituting ix and 'ix
into expression (1), we obtain:
>−
≤−=
WC
x
WC
xWxxxm
j
jj
jii
, C21
, 21
),( and
[ ]
−>−
−≤−−=
'
''
''
, 21
, 21
),(
ij
ijij
jii
xWC
xC
xWC
xxxWxxm . Comparing the two payoff functions, it is
clear that: ),(21
21
21
)( '' ji'iijjjiij xxmWxWxWx,xxmWC
x =+−>−=≤∀ ;
),(21
21
21
),( , '''' jiiijjiiij xxmWxWxCxxmxWC
WC
x =+−>−=
−∈∀ (since
CWxWC
x jj >⇒> ); and CxxmxxmxWC
x ji'ijiiij 21
),(),( - '' −==>∀ . This
completes stage 1 of the proof.
33 The rest of the proofs also focus on the American rule. Therefore, we omit the superscript A in the
payoff functions for notational convenience. 34 For the present purpose the level of optimism may be viewed as a strategy, and standard evolutionary
game theory may be used.
39
Stage 2: Unlike strictly dominated strategies or biases, weakly dominated strategies or
biases do not necessarily vanish in a payoff monotonic dynamic selection process.
However, if a strategy 'ix is weakly dominated by a strategy ix , and the subpopulation
programmed to strategy 'ix does not vanish, then the entire set of strategies against which
ix is better than 'ix vanish from the population (Weibull (1995), proposition 3.2). The
implication of this result in the present context is that either all pessimists vanish from the
population or that only pessimists and excessively optimistic players, with WC
x > ,
survive. The latter possibility is clearly unstable, since the excessively optimistic players
will earn a strictly higher payoff than the pessimistic players, and thus the pessimists will
be driven to extinction. Therefore, at the evolutionary equilibrium - 0)( 0 =<∀ xgx .
QED
Proof of Proposition 2: Part (i) of the proposition is proved below. The proof of part (ii)
follows a similar logic. The proof consists of two stages, similar to the two stages
employed in the proof of proposition 1. Stage 1 shows that excessive optimism is weakly
dominated by realism. Stage 2 utilizes the result of stage 1 to establish the disappearance
of excessive optimism through the evolutionary selection process.
Stage 1: Stage 1 proves that 0=ix weakly dominates WC
x 'i > . Substituting ix and 'ix
into expression (1), we obtain:
>−
≤−=
WC
x
WC
xWxxxm
j
jj
jii
, C21
, 21
),( and
40
[ ] ( )
−>−
<−≤−=
'
''
''
, 21
0 , 21
),(
ij
ijji
jii
xWC
xC
xWC
xxxWxxm . Since by proposition 1, all players with
0<jx will eventually vanish, we only need to compare the two payoff functions with
respect to 0≥jx . Through this comparison we obtain:
),(21
21
),( ' ji'ijjiij xxmCWxxxmWC
x =−>−=<∀ (since CWxWC
x jj <⇒< ); and
CxxmxxmWC
x ji'ijiij 21
),(),( ' −==≥∀ . This completes stage 1 of the proof.
Stage 2: Using the same reasoning employed in the proof of proposition 1, the weak
dominance of excessive optimism means that either only cautiously optimistic players,
with WC
x < , survive, or only excessively optimistic players, with WC
x > , survive. The
latter possibility is clearly unstable, since any mutant with WC
x2
≤ can successfully
invade such a population. Therefore, at the evolutionary equilibrium - 0)( =>∀ xgWC
x .
QED
Proof of Lemma 1: Part (i) of the lemma is proved below. The proof of part (ii) follows a
similar logic. To prove the evolutionary stability of WC
x̂2
= we have to show that no
mutant x̂xm ≠ can successfully invade the homogenous WC
x̂2
= -population. 35 This is
35 This condition is equivalent to the definition of an Evolutionary Stable Strategy (ESS) – See Maynard
Smith (1982) and Weibull (1995).
41
demonstrated through a comparison between the payoff obtained by a member of the
population and the payoff obtained by any possible mutant. Substituting the relevant
levels of optimism into expression (1) yields: 0)ˆ,ˆ( =xxmi ;
[ ] 021
<−=<∀ x̂xW)x̂,x(m x̂x mmmm ; and 021
<−=>∀ C)x̂,x(m x̂x mmm . Since any
mutant earns a strictly lower payoff compared to the homogenous population, WC
x̂2
= is
evolutionary stable. QED
Proof of Lemma 2: Part (i) of the lemma is proved below. The proof of part (ii) follows a
similar logic. Assume a homogenous population, in which the unique level of optimism is
WC
x̂2
> . We now show that a mutant WC
xm 2= can successfully invade the population.
This is demonstrated through a comparison between the payoff obtained by a member of
the population and the payoff obtained by the mutant. Substituting the relevant levels of
optimism into expression (1) yields: )ˆ,(21
)ˆ,ˆ( xxmCxxm mmi =−= and
),(021
),ˆ( mmmmi xxmCxxm =<−= , i.e. when matched against x̂ , both the population,
x̂ , and the mutant, mx , receive the same payoff, but when matched against mx , the
mutant receives a strictly higher payoff.
Next, examine the complementary case in which the unique level of optimism is
WC
x̂2
< . Again, a mutant WC
xm 2= can successfully invade the population. We
42
substitute the relevant levels of optimism into expression (1), and compare payoffs
0)ˆ,ˆ( =xxmi and [ ] 022
121
>
−=−= x̂
WC
Wx̂xW)x̂,x(m mmm (since WC
x̂2
< ). QED
43
Appendix B: The Numeric Algorithm
The analytical results presented in sections 3 and 4 have been supported and
extended using a computer based numeric algorithm. This algorithm has also been
applied to generate the results presented in section 5. Appendix B describes the numeric
algorithm and presents its implementation in the MATLAB 6 programming language.
The appendix also presents illustrative results of the numeric simulations.
Flow Chart
Figure 1a presents a flow chart of the evolutionary process studied in the present
paper. The legal setting is represented by the payoff function ),( jii xxm . The dynamics
are defined by the Dyn: )()( 1 xgxg tt +→ function, and the initial conditions of the
dynamic process are summarized in the initial distribution function, )(0 xg .
The algorithm consists of two main blocks. The first block calculates the fitness for
each optimism level. The second block embodies the dynamic structure vis-a-vis the
Dyn: )()( 1 xgxg tt +→ function. The algorithm applies a payoff monotonic dynamic
process. Hence, the output of the first block directly influences the dynamics of the
second block. The two procedures are carried out T times, where T represents the number
of periods specified for the dynamic process.
44
START
)x(g0 ; )x,x(m jii ; T
Dyn: )x(g)x(g tt 1+→
Calculate the fitness function ])x(g)x(m[E)x(f tt =
)]x(f),x(g[Dyn)x(g ttt =+1
t < T yes
END
no
Fig. 1a: A flow chart representation of the dynamic process
45
Implementation of the Algorithm in the MATLAB 6 Environment
The computer code, written in the MATLAB 6 programming language, is presented
below. The code is composed of three segments: the main program, the fitness function
(first block in the flow chart) and the Dyn function (second block in the flow chart). The
presented code implements the algorithm for a deterministic model of the American
regime. Implementation for the British rule requires minor adjustments. The simple type
of uncertainty analyzed in section 4 can be easily added through an appropriate
modification of the fitness function. Incorporating more complex forms of uncertainty
requires the application of a Monte Carlo procedure. The Monte Carlo technique
generates random errors (according to a given error distribution) that are then inserted
into the computer simulation, according to the specified information structure.
% **************** % The Main Program % **************** % Define System Parameters % ********************** global p C W; T = 50; % number of periods in the dynamic / evolutionary process % "Legal" Parameters % **************** p = 0.5; % probability of plaintiff victory at trial W = 500; % magnitude of judgment C = 100; % total litigation costs: C = Cp + Cd
46
% The Initial Distribution of Optimism in the Population % ******************************************** x_max = (p >= 0.5)*(1-p) + (p < 0.5)*p; % maximal level of optimism x_min = -x_max; % minimal level of optimism precision = 100; % number of optimism levels dx = (x_max - x_min)/precision; % size of optimism interval x = [x_min:dx:x_max]; [x_rows x_columns] = size(x); % A Uniform Distribution g0 = 1/(x_max - x_min)*ones(size(x)); % pdf % A Normal Distribution g0 = normpdf(x,0,x_max/5); % pdf % ******************************************** % The Dynamic Evolution of the Population % ********************************** g = zeros(T,x_columns); % initialize matrix of distributions fitness = zeros(T,x_columns); % initialize matrix of fitness values g(1,1:1:x_columns) = g0; % row 1 = Initial distribution for i=1:T fitness(i,1:1:x_columns) = f(g(i,1:1:x_columns),x); g(i+1,1:1 :x_columns) = dyn(g(i,1:1:x_columns),fitness(i,1:1:x_columns),dx);
end
47
% *************************** % ** The Fitness Function - f(g) ** % *************************** % % Description: The f(g) function calculates the fitness of every optimism level given the distribution of optimism in the population % ****************************************************************** function [f_res] = f(g,x) global p C W; x_min = min(x); x_max = max(x); dx = (x_max - x_min)/(length(x)-1); for i=1:length(x), for j=1:length(x), xpx(i,j) = x(i) + x(j); % calculate the (xi + xj) matrix xmx(i,j) = x(i) - x(j); % calculate the (xi - xj) matrix end end cond = ( xpx <= C/W ); m = 0.5*xmx*W.*cond – 0.5*C*(1-cond); f_res = (m*g')'; % ******************** % ** Function dyn(g,f) ** % ******************** % % Description: The Dyn(g,f) function defines the evolutionary dynamics. It calculates the fitness at period t+1 given the distribution of optimism and the fitness at period t % ******************************************************************* function [new_g] = dyn(g,f,dx) f_positive = f - min(f); new_g = g.*f_positive; new_g = new_g / sum(new_g)*dx;
48
The Numeric Calculations – Illustrative Results
Figure 2a demonstrates the convergence of a heterogeneous population (with
]5.0,5.0[~ −Ux ) to the unique level of optimism, WC
2, in the deterministic model of an
American regime. As explained in section 3, the pessimists disappear first (see
proposition 1). While pessimists exist, the excessive optimists prosper by exploiting the
pessimists. However, the pessimists soon disappear, and then the excessive optimists are
also driven out of the market (see proposition 2). Finally, the population converges to the
unique level of optimism, WC
2.
49
Fig. 2a: Convergence to the evolutionary stable level of optimism
50
References
Alchian, A. A. (1950), "Uncertainty, Evolution and Economic Theory", Journal of
Political Economics, 58, 211-221.
Bar-Gill, O. (2000), “The Success and Survival of Cautious Optimism: An Evolutionary
Analysis of Pretrial Settlement Negotiations”, The Eitan Berglas School of
Economics (Tel-Aviv University) Working Paper Series, WP No. 15-2000 (June,
2000).
Bar-Gill, O. (2002), “Overconfidence, Optimism and Market Contracts: Cognitive Biases
in the Shadow of the Law”, mimeo, Harvard Law School.
Bar-Gill, O. and Fershtman, C. (2000), “The Limit of Public Policy: Endogenous
Preferences”, Tilburg University, CentER Discussion Paper #2000-71.
Bebchuk, L. A. (1984), "Litigation and Settlement under Imperfect Information", RAND
Journal of Economics, 15, 404-415.
Bester, H. and Guth, W. (1998), "Is Altruism Evolutionary Stable ?", Journal of
Economic Behavior and Organization, 34, 211-221.
Birke, R. and Fox, C. R. (1999), "Psychological Principals in Negotiating Civil
Settlements", Harvard Negotiations Law Review, 4, 1-57.
Bottom, W. P. and Paese, P. W. (1999), "Judgment Accuracy and the Asymmetric Cost
of Errors in Distributive Bargaining", Group Decision and Negotiation, 8, 349-364.
Bowles, S. (1998), "Endogenous Preferences: The Cultural Consequences of Markets and
other Economic Institutions", Journal of Economic Literature, 36, 75-111.
Camerer, C. (1992), "The Rationality of Prices and Volume in Experimental Markets",
Organizational Behavior and Human Decision Processes, 51, 237-272.
51
Cooter, R., Marks, S. and Mnookin, R. (1982), "Bargaining in the Shadow of the Law: A
Testable Model of Strategic Behavior", Journal of Legal Studies, 11, 225-251.
Cooter, R. and Ulen, T. (1999), Law and Economics (3rd edition, Reading, MA, Addison-
Wesley, Inc.).
Daughety, A. F. (2000), “Settlement”, in Bouckaert, B. and De Geest, G. (eds.),
Encyclopedia of Law and Economics, Vol. 5, pp. 95-158 (Northhampton, MA,
Edward Elgar Publishing Co.).
Dekel, E. and Scotchmer, S. (1999), "On the Evolution of Attitudes toward Risk in
Winner-Take-All Games", Journal of Economic Theory, 87, 125-143.
Dutta, P. K. and Radner, R. (1999), “Profit Maximization and the Market Selection
Hypothesis”, Review of Economic Studies, 66, 769-798.
Ellingsen, T. (1997), "The Evolution of Bargaining Behavior", Quarterly Journal of
Economics, 112, 581-602.
Farmer, A and Pecorino, P. (1996), “Issues of Informational Asymmetry in Legal
Bargaining”, in Anderson, D. A. (Ed.), Dispute Resolution: Bridging the Settlement
Gap (Greenwich, CT, JAI Press).
Farmer, A and Pecorino, P. (2000), “Pretrial Settlement with Fairness”, The University of
Alabama Economics, Finance and Legal Studies Working Paper Series, Working
Paper 00-10-03.
Fershtman, C. and Judd, K. L. (1987), “Equilibrium Incentives in Oligopoly”, American
Economic Review, 77, 927-940.
52
Fershtman, H. and Weiss, Y. (1997), "Why Do We Care About What Others Think about
Us", in Ben-Ner, A. and Putterman, L. (Eds.), Economics, Values and Organization
(Cambridge, UK, Cambridge University Press).
Fershtman, H. and Weiss, Y. (1998), "Social Rewards, Externalities and Stable
Preferences", Journal of Public Economics, 70, 53-74.
Frank, R. H. (1988), Passion with Reason - The Strategic Role of the Emotions (New
York, NY, W.W. Norton & Company).
Friedman, M. (1953), Essays in Positive Economics (Chicago, IL, University of Chicago
Press).
Gilson, R. J. and Mnookin, R. H. (1994), “Disputing Through Agents: Cooperation and
Conflict Between Lawyers in Litigation”, Columbia Law Review, 94, 509-566.
Gould, J. P. (1973), "The Economics of Legal Conflict", Journal of Legal Studies, 2, 279-
300.
Dufwenberg M. and Güth, W. (1999), “Indirect evolution vs. strategic delega tion: a
comparison of two approaches to explaining economic institutions”, European
Journal of Political Economy, 15, 281-295.
Güth, W. and Yaari, M. E. (1992), "Explaining Reciprocal Behavior in Simple Strategic
Games: An Evolutionary Approach", in Witt U. (Ed.), Explaining Process and
Change: Approaches to Evolutionary Economics (Ann Arbor, MI, University of
Michigan Press).
Hay, B. and Spier, K. E. (1998), “Settlement of Litigation”, in Newman P. (ed.), “The
New Palgrave Dictionary of Economics and the Law, pp. 442-451 (Gordonsville,
VA, Palgrave/St. Martin’s Press).
53
Heifetz A. and Segev, E. (2001), “The Evolutionary Role of Toughness in Bargaining”,
mimeo, Tel-Aviv University, Tel-Aviv.
Heifetz, A. and Spiegel, Y. (1999), "On the Evolutionary Emergence of Optimism",
mimeo, Tel-Aviv University, Tel-Aviv.
Huck, S., Kirchsteiger, G. and Oechssler, J. (1997), "Learning To Like What You Have –
Explaining The Endowment Effect", mimeo, Humboldt University, Berlin.
Huck, S. and Oechssler, J. (1998), "The Indirect Evolutionary Approach to Explaining
Fair Allocations", Games and Economic Behavior, 28, 13-24.
Kagan, R. A. (1994), “Do Lawyers Cause Adversarial Legalism? A Preliminary Inquiry”,
Law and Social Inquiry, 19, 1-62.
Kaplow, L. and Shavell, S. (2002), “Economic Analysis of Law”, in Auerbach, A. J. and
Feldstein, M. S. (Eds.), Handbook of Public Economics, Vol. 3, ch. 25.
Kyle, A. S. and Wang, F. A. (1997), "Speculation Duopoly with Agreement to Disagree:
Can Overconfidence Survive the Market Test?", The Journal of Finance, LII, 2073-
2090.
Landes, W. M. (1971), "An Economic Analysis of the Courts", Journal of Law and
Economics, 14, 61-107.
Lowenstein, G., Issacharoff, S., Camerer, C. and Babcock, L. (1993), "Self-Serving
Assessments of Fairness and Pre-Trial Bargaining", Journal of Legal Studies, 22,
135-159.
Majumdar, M. and Radner, R. (1991), “Linear Models of Economic Survival under
Production Uncertainty”, Economic Theory, 1, 13-30.
54
Maynard Smith, J. (1982), Evolution and the Theory of Games (Cambridge, UK,
Cambridge University Press).
Nalebuff, B. (1987), "Credible Pretrial Negotiations", RAND Journal of Economics, 18,
198-210.
Nash, J. F. (1950), “The Bargaining Problem”, Econometrica, 18, 155-162.
Osiel, M. J. (1990), “Lawyers as Monopolists, Aristocrats and Entrepreneurs, Review of
Lawyers in Society, edited by Abel, R. L. and Lewis, P. S. C. (1988)”, Harvard Law
Review, 103, 2009-2066.
Posner, R. A. (1973), "An Economic Approach to Legal Procedure and Judicial
Administration", Journal of Legal Studies, 2, 399-458.
Posner, R. A. (1998), Economic Analysis of Law (5th edition, New York, NY, Aspen
Publishers, Inc.).
Priest, G. L. and Klein, B. (1984), "The Selection of Disputes for Litigation", Journal of
Legal Studies, 13, 1-55.
Reinganum, J. F. and Wilde, L. L. (1986), "Settlement, Litigation and the Allocation of
Litigation Costs", RAND Journal of Economics, 17, 557-566.
Shavell, S. (1982), “Suit, Settlement, and Trial: A Theoretical Analysis under Alternative
Methods for the Allocation of Legal Costs”, Journal of Legal Studies, 11, 55-81.
Spier K. E. (1992), “The Dynamics of Pretrial Negotiations”, Review of Economic
Studies, 59, 93-108.
Spier, K. E. (1994), “Pretrial Bargaining and the Design of Fee Shifting Rules”, Rand
Journal of Economics, 25, 197-215.
55
Waldman, M. (1994), "Systematic Errors and the Theory of Natural Selection", American
Economic Review, 84, 482-497.
Weibull, J. W. (1995), Evolutionary Game Theory (Cambridge, MA, MIT Press).
Young, H. P. (1993), "An Evolutionary Model of Bargaining", Journal of Economic
Theory, 59, 145-168.